Super-polynomial accuracy of multidimensional randomized nets using the median-of-means

Abstract
We study approximate integration of a function over based on taking the median of integral estimates derived from independently randomized -nets in base . The nets are randomized by Matousek's random linear scramble with a digital shift. If is analytic over , then the probability that any one randomized net's estimate has an error larger than times a quantity depending on is for any . As a result the median of the distribution of these scrambled nets has an error that is for function evaluations. The sample median of independent draws attains this rate too, so long as is bounded away from zero as . We include results for finite precision estimates and some non-asymptotic comparisons to taking the mean of independent draws.
View on arXivComments on this paper